Abstract
With the rapid development of high-speed image sensors and optical imaging technology, these have effectively promoted the improvement of non-contact 3D shape measurement. Among them, striped structured-light technology has been widely used because of its high measurement accuracy. Compared with classical methods such as Fourier transform profilometry, many deep neural networks are utilized to restore 3D shape from single-shot structured light. In actual engineering deployments, the number of learnable parameters of convolution neural network (CNN) is huge, especially for high-resolution structured-light patterns. To this end, we proposed a dual-path hybrid network based on UNet, which eliminates the deepest convolution layers to reduce the number of learnable parameters, and a swin transformer path is additionally built on the decoder to improve the global perception of this network. The experimental results show that the learnable parameters of the model are reduced by 60% compared with the UNet, and the measurement accuracy is not degraded at the same time. The proposed dual-path hybrid network provides an effective solution for structured-light 3D reconstruction and its practice in engineering.
Highlights
As a high-accuracy non-contact 3D reconstruction method, the striped structured-light profilometry has been widely employed in geometric measurement, relic restoration, reverse engineering, etc. [1, 2]
In the structure comparing of Hybrid-A and Hybrid-B, we find out that if the deep feature map is extracted only once from the convolution path for global feature representation, even if the global features are fed back to the convolution path for many times, the back-propagation of the network still has the problem of gradient disappearance, which leads to the weak generalization capability of the hybrid network, and the decline of the accuracy of 3D reconstruction
In Hybrid-A, the swin transformer path can extract local feature information from the convolution path repeatedly and fuses them with the global features from the previous swin transformer layer, just like the residual skip connection of ResNet, which can avoid the gradient disappearance of the model and help to improve the 3D reconstruction accuracy [52, 53]
Summary
As a high-accuracy non-contact 3D reconstruction method, the striped structured-light profilometry has been widely employed in geometric measurement, relic restoration, reverse engineering, etc. [1, 2]. As a high-accuracy non-contact 3D reconstruction method, the striped structured-light profilometry has been widely employed in geometric measurement, relic restoration, reverse engineering, etc. Structured light measurement has been applied in automatic driving in recent years, while the real-time perception of vehicles and obstacles is the research hotspot of IOV (Internet of vehicle) and an important prerequisite for automatic and unmanned driving [3]. LIDAR has a long detection distance, but the resolution of 3D imaging is low, and the high cost restricts its application, while a single camera without auxiliary lighting cannot calculate the 3D profile of the road environment [4, 5]. With the help of invisible light sources, a single camera can carry out road environment reconstruction based on the structuredlight profilometry.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
